Reproduce “Spanning Diffusion and Flow-based Posterior Sampling via Exact Gaussian Conditioning” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Teo Deveney, Lea Bogensperger, Hok Shing Wong
β-Flow restores images from incomplete measurements by combining a denoiser’s clean-image estimate with a Gaussian measurement correction, while controlling correlation between successive trajectory states and prior uncertainty. In AFHQ-Cat 6× super-resolution, at β=0.6, κ=0.1 yields 25.57 dB PSNR and 0.712 SSIM, versus 20.66 dB and 0.553 at κ=3; higher scores are better, and this comparison is task-specific.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.